Qwen3
The latest large language model in the Qwen series, with publicly available weights and both dense and Mixture-of-Experts (MoE) architectures.
About Qwen3
Qwen3 is the newest addition to the Qwen family of large language models, developed by the Qwen team (Alibaba Cloud). It builds on the experience gained from QwQ and Qwen2.5, making it the more advanced and capable model in the series to date. Qwen3's weights are publicly available, including both dense and Mixture-of-Experts (MoE) models. All open-weight models are released under the Apache 2.0 license.
Best for
Best use cases
Deploying an open-source large language model locally or on your own infrastructure
Enabling thinking mode when handling complex logical reasoning, math, and programming tasks
Switching to non-thinking mode for efficient general conversation
Developing or researching on top of the model under the Apache 2.0 license
Features
Multiple model sizes and architectures
Offers both dense and Mixture-of-Experts (MoE) models, with parameter sizes covering 0.6B, 1.7B, 4B, 8B, 14B, 32B, 30B-A3B, and 235B-A22B.
Thinking and non-thinking mode switching
Seamlessly switch between thinking mode (for complex logical reasoning, math, and programming) and non-thinking mode (for efficient general conversation).
Enhanced reasoning capability
Reasoning capability surpasses the previous QwQ in thinking mode and Qwen2.5 instruct models in non-thinking mode.
Apache 2.0 licensing
All open-weight models are released under the Apache 2.0 license.
Pros
- Publicly available weights covering both dense and MoE model types
- Multiple parameter sizes available: 0.6B, 1.7B, 4B, 8B, 14B, 32B, 30B-A3B, and 235B-A22B
- Seamless switching between thinking and non-thinking modes
- Reasoning capability surpasses the previous QwQ in thinking mode and Qwen2.5 instruct models in non-thinking mode
- All open-weight models are released under the Apache 2.0 license